1

131. Numeral Bases

Bernard Comrie

1.

Introduction

This map is concerned with one aspect of the mathematical
structure of linguistic expressions for numerals, namely the
arithmetic base that is used in constructing numeral
expressions. By the “base” of a numeral system we mean the
value

n

such that numeral expressions are constructed

according to the pattern

... xn + y

, i.e. some numeral

x

multiplied by the base plus some other numeral. (The order of
elements is irrelevant, as are the particular conventions used in
individual languages to indicate multiplication and addition.) A
simple example is provided by Mandarin, with base 10, in which
the numeral 26 is expressed as in (1).

(1) Mandarin

èr-shí-lìu

two-ten-six

In Mandarin, the convention is that the numeral before the word
for 10 is to be multiplied by 10, while that after the word for 10
is to be added to this product ([2 x 10] + 6). Using this concept
of base, plus some additional concepts to be introduced below,
six main numeral systems can be identified, of which the second
and third in the feature value box can be viewed as subtypes of
one superordinate type.

@ 1. Decimal

125

@ 2. Hybrid

vigesimal–decimal

22

@ 3. Pure

vigesimal

20

@ 4. Other

base

5

@ 5. Extended

body-part

system

4

2

@ 6. Restricted

20

total 196

As the Mandarin example shows, the crucial concepts

needed to demonstrate that a numeral system has a particular
base are addition and multiplication. Beyond this, many numeral
systems also make use of exponentiation of the base, i.e.
expressions to denote the result of raising the base to various
powers. Thus, English has a decimal system, and has a special
term for 10

2

, namely

hundred

, as well as one for 10

3

, namely

thousand

. While the use of exponentiation often reinforces the

identification of the base, it is not taken here as a defining
feature, since some languages use addition and multiplication
but without making use of exponentiation; an example is
Chukchi (Chukotko-Kamchatkan; eastern Siberia), with a
vigesimal (base 20) system, but no special expression for 20

2

,

i.e. 400, which is simply expressed as 20 x 20. Moreover, the
linguistic expression of exponentiation is often opaque — there
is nothing in the form of the English words

hundred

and

thousand

to indicate that they are, respectively, the second and

third powers of the base. Even the limited transparency provided
by numerals like English

bi-llion

,

tri-llion

, with Latinate prefixes

for 2 and 3 respectively, is only related by a quite complex
formula to the corresponding power of the base 10: 10

3(n + 1)

in

American usage or 10

6n

in traditional British usage. For further

general discussion, see Comrie (1997), Greenberg (1978), and
Hurford (1975).

2. The

six

types

The decimal system has already been introduced by means of
the Mandarin example (1); the general structure of numerals in a
decimal system is

x10 + y

.

In a pure vigesimal system, the base is consistently 20,

i.e. the general formula for constructing numerals is

x20 + y

. An

3

example is provided in (2) by Diola-Fogny (Atlantic, Niger-
Congo; Senegal), in which the numeral 51 is expressed as ‘two
twenties and eleven’.

(2)

Diola-Fogny (Sapir 1965: 84–85)

bukan ku-gaba di uLMn di b-NkOn

twenty

CL

6-two and ten and

CL

9-one

For practical reasons — in particular, the frequency of the

type in the world’s languages — it is useful to distinguish a
hybrid vigesimal–decimal system in which the numbers up to 99
are expressed vigesimally, but the system then shifts to being
decimal for the expression of the hundreds, so that one ends up
with expressions of the type

x100 + y20 + z

; this is illustrated

in (3) by the Basque expression for 256:

(3) Basque (Oroitz Jauregi, p.c.)

berr-eun eta

berr-ogei-ta-hama-sei

two-hundred

and

two-twenty-and-ten-six

Bases other than 10 and 20 are also attested, albeit rarely,

among the world’s languages. Ekari (Trans-New Guinea; Papua,
Indonesia) makes use of a base of 60, as illustrated in the
expression for 71 in (4); the base of 60 was also used in the
ancient Near Eastern language Sumerian.

(4) Ekari (Drabbe 1952: 30)

èna ma gàati dàimita mutò

one

and

ten

and sixty

Some languages of the world have numeral systems that

do not make use of an arithmetic base. One such system is the
extended body-part system, here illustrated by a discussion of
Kobon (Madang, Trans-New Guinea), which is quite typical of a
number of languages of Highland New Guinea. Languages like

4

Kobon make use of further body parts to extend the system
beyond the ten fingers. In Kobon specifically, the names of the
following body parts (on the left-hand side of the body) are
used in order to count from 1 to 12: little finger, ring finger,
middle finger, index finger, thumb, wrist, middle of forearm,
inside of elbow, middle of upper arm, shoulder, collarbone, hole
above breastbone. The count can then continue down the right-
hand side of the body, from the collarbone to the (right)
shoulder as 13 to the little finger as 23. It is then possible to
reverse the count, starting from the little finger of the right hand
as 24 back up to the hole above the breastbone as 35 and down
again to the little finger of the left hand as 46. One effect of this
is that the names of particular body parts when used as
numerals are multiply ambiguous. For instance,

siduT

‘shoulder’

can denote either 10 (on the left-hand side of the first pass
across the body), or 14 (on the right-hand side of that pass), or
33 (on the right-hand side of the return pass across the body),
or 37 (on the left-hand side of that pass), or 56 on the left-hand
side of the next pass across the body, etc. There are usually
means, optional or obligatory depending on the language, to
distinguish the second side of the body used in a count from the
first, as well as to indicate which pass across the body is being
used, but there is no productive means to identify other than a
small number of passes across the body. Extended body-part
systems are thus typically rather limited in the range of numbers
that they can express, but can be used productively at least into
the scores.

Finally, some languages have restricted numeral systems,

by which I mean more specifically a numeral system that does
not effectively go above around 20. The most restricted numeral
system would of course be one lacking any numerals at all, and
according to Dan Everett (personal communication) Pirahã
(Mura; Brazil) is a language of just this type. A number of
languages of the world have numeral systems that extend only
as far as 3 (e.g. Mangarrayi (isolate; Northern Territory,

5

Australia)), while others show slightly higher but nonetheless
heavily restricted upper limits, such as 5 (Yidiny (Pama-
Nyungan; Queensland, Australia)).

3. Problem

cases

In many cases, the assignment of a language to one or another
of the types identified is straightforward, but nonetheless a
number of problems can arise, and the following paragraphs will
note some of these and, where relevant, indicate the solutions
that have been adopted in the data analysis underlying this
chapter.

First, it is essential to ascertain that the expressions in

question are indeed numerals, since in many languages there
are quantifying expressions, including some with quite specific
denotations, other than numerals, such as

pair

in English

(necessarily denoting a set of 2). The general criterion to be
used is that for an expression to qualify as a numeral, it must be
the usual way of identifying that number of entities in the
language in question within a noun phrase. In modern standard
English,

seventy

(as in

seventy years

) is thus a numeral, whereas

three score and ten

is not, even if there may have been periods

in the history of the English language, and may still be regional
varieties, where the latter expression would qualify as a
numeral. Note that it is probably not reasonable to require that
numerals be used in counting — some cultures with low
numeracy do not engage in counting, although they may
nonetheless have a non-empty restricted numeral system.

Some languages have two (or more) numeral systems

satisfying the criterion of the previous paragraph. Where both
are of the same type, as with the indigenous and Sino-Korean
systems in Korean — both decimal — then there is no problem
in assigning the language unequivocally to one type. Where the
systems are of different types, preference has been given to the

6

most productive, e.g. the extended body-part system in Kobon,
which exists alongside a restricted system.

Many languages combine different bases in the

construction of their numeral system, and for the purposes of
this chapter various decisions have been taken, some principled,
some of a more practical nature, to limit the number of distinct
types represented on the map. Only one mixed-base type has
been given a separate representation, namely the type that is
vigesimal in the range up to 99 and then decimal in the
expression of the hundreds, because of its frequent occurrence
among the languages of the world. In the case of other mixed
systems, preference has been given to the base that is most
productive in the construction of numerals in the range 20–400.
In all numeral systems with a base of 20 or greater, and in
several with a smaller base, numbers less than the base are
constructed using smaller bases. For instance, in Igbo (Niger-
Congo; Nigeria), with a base of 20, the numerals 1-19 are
constructed using 10 as the base, as illustrated in (5) for 32:

(5)

Igbo (Green and Igwe 1963: 37)

ohu nà ìri nà àbuYòY

twenty and ten and two

In Supyire (Gur, Niger-Congo; Mali), with a base of 80, the lower
scores are expressed vigesimally, while numbers below 20 are
expressed using a mixed quinary–decimal system (base 5), as
illustrated in (6), with 399 expressed as ’80 x 4 + 20 x 3 + 10
+ 5 + 4’:

(6) Supyire (Carlson 1994: 169)

TZkwuu sicyMMré \ná béé-tàànrè

eighty

four and

twenty-three

ná kM^

\ná báárì-cyMZMZrè

and ten and five-four

7

In some cases, an alternative base is used only in the
construction of a small proportion of relevant numerals, and in
such cases this alternative is simply disregarded in assigning the
numeral system overall to a particular type. In French, for
instance, the numerals in the range 80–99 have a vigesimal
structure, as in the expression for 97 in (7), but since the
system is otherwise entirely or almost entirely decimal, French
has been assigned to the decimal type.

(7) French

quatre-vingt-dix-sept

four-twenty-ten-seven

While some languages, like Mandarin illustrated in (1),

have completely or almost completely transparent structures of
the type

xn + y

, other languages have various departures from

the ideal formula in their numeral expressions through the
appearance of morphophonological idiosyncrasies or
portmanteau forms, and it is clearly necessary to be able to
distinguish such exceptions from an instantiation of a different
numeral system type. Relatively minor morphophonological
idiosyncrasies can usually be identified without difficulty, as in
relating English

fif-ty

to

five

and

ten

, with the identification of

the first morpheme more transparent than that of the second. In
some cases, completely different phonological forms may be
used in certain combinations, as in the expression of the tens in
Spanish, where the suffix

–enta

, as in

och-enta

80 (eight-ten),

bears no formal resemblance to

diez

10, but is nonetheless

reasonably consistent in the expression of the tens. When only a
handful of forms are portmanteau in an otherwise transparent
system, they can be disregarded, as in the case of Russian
monomorphemic

sorok

40 in comparison with

pjat´-desjat

50

(five-ten). In some languages most or all of the products of the
base are portmanteau forms, but we still identify them as such
in that there is a separate such form for each product of the

8

base and intermediate numbers are expressed by adding to that
form; thus, in Turkish (Turkic; Turkey) each of the tens in the
range 20–50 is monomorphemic (

yirmi

20,

otuz

30,

k`rk

40,

elli

50; cf.

iki

2,

üç

3,

dört

4,

bed

5), but numbers between the tens

are expressed by addition of the remainder, as in (8), which
expresses 21:

(8) Turkish (Kornfilt 1997: 428)

yirmi bir

twenty

one

4. Geographical

distribution

Even a cursory perusal of the accompanying map serves to show
that, at least as far as numeral systems are concerned, we live in
a decimal world, with the decimal type dominant in nearly every
part of the world. Some of the other types are highly restricted
geographically, in particular the extended body-part type, found
as a basic numeral system in Highland New Guinea, and the
restricted system, largely confined to Australia and Amazonia.
Bases other than 10 or 20 are extremely rare in the modern
world, the examples in our sample being Supyire in West Africa
and Ekari in Indonesian Papua.

However, the vigesimal system, whether pure or combined

with the decimal system above 100, is still found in a number of
different areas in the world, and is particularly frequent in some
specific areas, such as Mesoamerica. Mesoamerica is in fact
indicative of a worldwide historical trend for the dominant
decimal system to encroach on and replace other systems. Pre-
Conquest Mesoamerica was largely vigesimal, with the
prototypical example being Classical Mayan. The influence of
Spanish after the Conquest led to many indigenous languages
adopting the Spanish numeral

ciento

100 and with it the decimal

system for the expression of the hundreds. In many languages,
replacement by Spanish forms has percolated even further down

9

the system, and it is not infrequent in contemporary accounts of
Mesoamerican languages to read that in practice Spanish
numeral expressions are used for all but the lowest numbers.
Non-decimal numeral systems are even more endangered than
the languages in which they occur.